BUG-Performance of tp.create_scenario about taipy-core

[florian-vuillemot]

Pipeline manager:

@classmethod
    def _get_all_by_config_id(cls, config_id: str) -> List[Pipeline]:
        return cls._repository._search_by_name(config_id)

fs_base

    def _search_by_name(self, name) -> List[Entity]:
        return [self.__to_entity(f) for f in self._directory.glob(f"*{name}*.json")]

On pipeline, scenario, task and data node => 0.21 to 0.10 for latest insertion after 50.

Better:
return [self.__to_entity(f) for f in self._directory.iterdir() if name in f.name]

[florian-vuillemot]

__get_model_filepath in _fs_base.py

        if raise_if_not_exist and not filepath.exists():
            raise ModelNotFound(str(self._directory), model_id)

and not if not filepath.exists() and raise_if_not_exist:

[florian-vuillemot]

Impact time is on _search_by_name in _fs_base.

[florian-vuillemot]

do not reload config_id, parent_id, etc.

[FlorianJacta]

Here is the new performance with these changes. Note that for cycle, I create scenarios with the same cycle/day so at the end, I only have one cycle in my .data.

We can predict the time execution of create_scenario:

Time (no cycle) = 0.06735397nb_scenarios + 0.07187587
Time (cycle) = 0.14598272nb_scenarios + 0.037

It means that the coefs between no cycle and cycle is multiplied by 2,16739592. The intercept for both times are almost null.

Moreover, the coef of the last version was around 0.28 for no cycle. It means that time execution was divided by 4 for no cycle.
The coef for cycle was around 0.46. So the time was divided by 3.2.